Browsing by Author "Chen, T. P."

Now showing 1 - 5 of 5

A Novel Architecture to Build Ideal-linearity Neuromorphic Synapses on a Pure Logic FinFET Platform Featuring 2.5ns PGM-time and 10(12) Endurance
(2019-01-01)
Hsieh, E. R.
;
Chang, H. Y.
;
Chung, Steve S.
;
Chen, T. P.
;
Huang, S. A.
In this work, we will explore pure logic FinFET devices to realize the functionality of linear weight tuning capability as electric synapses. The unit cell of this new FinFET synapse is composed of two identical FinFETs in series; one serves as control and the other one as storage. This new FinFET synapse exhibits ideal linearity with nearly infinity training cycles (>10(12)). much lower programming voltage, 0.85V, and faster speed, 2.5ns. It can also analogically increase or decrease the transistor's V-th, to vary the drain conductance. As far as the analog performance is concerned, it performs excellent linearity and a wide tuning-window (20x) of weight-tuning capability. 1kb synaptic array has also been designed. The spice-simulated results have shown that new FinFET synaptic array can expand the array-size to 64x64, exhibiting 300x of SNR, w.r.t. that of RRAM array. Finally, the training of the neural network based on the proposed FinFET synapse can achieve 97.43% accuracy as high as the GPU one does.
An Energy Efficient FinFET-based Field Programmable Synapse Array (FPSA) Feasible for One-shot Learning on EDGE AI
(2018-01-01)
Kuo, J. L.
;
Chen, H. W.
;
Hsieh, E. R.
;
Chung, Steve S.
;
Chen, T. P.
A pure logic 14nm FinFET with capabilities of linearly tunable V-th and excellent retention has been implemented as synapses in neuromorphic system., a Field Programmable Synapse Array (FPSA) has been adopted to replace conventional R-based memory Synapse Array (RSA). Thanks to the wide range of V-t-tuning ability, 200X on/off ratio, and the ultra-small variability, 12%, results showed that the training power and SN ratio of FPSA are 10 times and 50 times smaller than those of the RSA, respectively. Two applications were demonstrated on FPSA array for one-shot learning applications. First, FPSA is used to detect handwritten digits of MNIST dataset. "Learned it by once" can be achieved in this task. Furthermore, FPSA has been applied to recognize goldfish in Cifar 100 dataset after learned the other 4 fish species. With the assistance from one-shot learning, results show the machine learned it faster and better on EDGE. This demonstrates the feasibility of FPSA for low-power and cost-effective synapse-based one-shot learning applications in the AIoT era.
Embedded PUF on 14nm HKMG FinFET Platform: A Novel 2-bit-per-cell OTP-based Memory Feasible for IoT Secuirty Solution in 5G Era
(2019-01-01)
Hsieh, E. R.
;
Wang, H. W.
;
Liu, C. H.
;
Chung, Steve S.
;
Chen, T. P.
In this work, a novel concept of 2-bit-per-cell (2B/C) is introduced to realize high-density OTP PUF from a new scheme of dielectric breakdown. This PUF shows 10(5)x of large window, good immunity to high-temperature disturbance, and excellent retention under 150 degrees C baking, which are particularly for automotive applications. In terms of security. this PUF exhibits near ideal normal distribution of hamming distance and narrow distribution of hamming weight. The bit error rates are low, 0.78% at 25 degrees C and 1.95% at 150 degrees C, benchmarked on a 256-bit array. Finally, the security test of this PUF against the hackers' attack from the machine learning process has been proved to have high security. Overall, the proposed 2B/C OTP PUF demonstrated great potential for IoT security in 5G era.
Novel Concept of the Transistor Variation Directed Toward the Circuit Implementation of Physical Unclonable Function (PUF) and True-random-number Generator (TRNG)
(2019-01-01)
Xiao, Y.
;
Hsieh, E. R.
;
Chung, Steve S.
;
Chen, T. P.
;
Huang, S. A.
For the first time, we use a unique feature of S/D variation of FinFET to realize PUF (Physical unclonable function) and TRNG (True random number generator). With the scaling of transistors, S/D variation becomes significant. This provides an opportunity to use the mismatch of the S/D resistances for the design of PUF. Method has been developed to first rule out those less significant factors of variation. Then, a PUF is developed based on the dominant S/D variation. In terms of the security, this PUF exhibits ideal un-biased normal distribution of hamming distance, and narrow distribution of hamming weight in the range of 45%similar to 55%. The unstable bit rates are also very low (1.17%) under room temperature and 3.12% under 150 degrees C, benchmarked on a 256 bits array. Meanwhile, as a result of the defect (in the form of traps) at the drain/substrate junction, RTN behavior was observed from the current measured between drain/source and substrate, named I-b-RTN. It provides us a way to implement a TRNG. This TRNG passed NIST test up to 9 items. Overall, the S/D mismatch PUF and I-b-RTN TRNG demonstrated great potential to meet the requirements of the IoT security application.
Performance enhancement of GaN-based flip-chip ultraviolet light-emitting diodes with a RPD AlN nucleation layer on patterned sapphire substrate
(2014-08-01)
Lin, B. C.
;
Chiu, C. H.
;
Lee, C. Y.
;
Han, H. V.
;
Tu, P. M.
In this work, flip-chip ultraviolet light-emitting diodes (FCUV-LEDs) on patterned sapphire substrate (PSS) at 375 nm were grown by an atmospheric pressure metal organic chemical vapor deposition (AP-MOCVD). A specialized reactive plasma deposited (RPD) AlN nucleation layer was utilized on the PSS to enhance the quality of the epitaxial layer. By using high-resolution X-ray diffraction, the full-width at half-maximum of the rocking curve shows that the FCUV-LEDs with RPD AlN nucleation layer had better crystalline quality when compared to conventional GaN nucleation samples. From the transmission electron microscopy (TEM) image, it can be observed that the tip and incline portion of the pattern was smooth using the RPD AlN nucleation layer. The threading dislocation densities (TDDs) are reduced from 7 x 10(7) cm(-2) to 2.5 x 10(7) cm(-2) at the interface between the u-GaN layers for conventional and AlN PSS devices, respectively. As a result, a much higher light output power was achieved. The improvement of light output power at an injection current of 20 mA was enhanced by 30%. Further photoluminescence measurement and numerical simulation confirm such increase of output power can be attributed to the improvement of material quality and light extraction. (C) 2014 Optical Society of America